[PATCH 24/25] x86/mcheck: Move CPU_ONLINE to hotplug state machine

[Sebastian Andrzej Siewior]

This callback is still partly asymmetrical since the counterpart of
mce_device_create is done in CPU_DEAD.

On failure we don't undo mce_device_create() doing _but_ it will happen
once we move CPU_DEAD to the state machine.

Cc: Tony Luck <tony.luck@xxxxxxxxx>
Cc: Borislav Petkov <bp@xxxxxxxxx>
Cc: linux-edac@xxxxxxxxxxxxxxx
Cc: x86@xxxxxxxxxx
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@xxxxxxxxxxxxx>
Signed-off-by: Thomas Gleixner <tglx@xxxxxxxxxxxxx>
---
 arch/x86/kernel/cpu/mcheck/mce.c | 61 ++++++++++++++++------------------------
 1 file changed, 24 insertions(+), 37 deletions(-)

diff --git a/arch/x86/kernel/cpu/mcheck/mce.c b/arch/x86/kernel/cpu/mcheck/mce.c
index 596a7128a46b..b1770ebcb8de 100644
--- a/arch/x86/kernel/cpu/mcheck/mce.c
+++ b/arch/x86/kernel/cpu/mcheck/mce.c
@@ -2487,18 +2487,6 @@ mce_cpu_callback(struct notifier_block *nfb, unsigned long action, void *hcpu)
 	unsigned int cpu = (unsigned long)hcpu;
 
 	switch (action & ~CPU_TASKS_FROZEN) {
-	case CPU_ONLINE:
-		mce_device_create(cpu);
-		if (threshold_cpu_callback_online) {
-			int ret;
-
-			ret = threshold_cpu_callback_online(cpu);
-			if (ret) {
-				mce_device_remove(cpu);
-				return NOTIFY_BAD;
-			}
-		}
-		break;
 	case CPU_DEAD:
 		if (threshold_cpu_callback_dead)
 			threshold_cpu_callback_dead(cpu);
@@ -2514,6 +2502,22 @@ mce_cpu_callback(struct notifier_block *nfb, unsigned long action, void *hcpu)
 	return NOTIFY_OK;
 }
 
+static int mce_cpu_online(unsigned int cpu)
+{
+	int ret;
+
+	mce_device_create(cpu);
+	if (!threshold_cpu_callback_online)
+		return 0;
+
+	ret = threshold_cpu_callback_online(cpu);
+	if (ret) {
+		mce_device_remove(cpu);
+		return ret;
+	}
+	return 0;
+}
+
 static int mce_cpu_down_dying(unsigned int cpu)
 {
 	struct timer_list *t = this_cpu_ptr(&mce_timer);
@@ -2547,8 +2551,8 @@ static __init void mce_init_banks(void)
 
 static __init int mcheck_init_device(void)
 {
+	enum cpuhp_state hp_online;
 	int err;
-	int i = 0;
 
 	if (!mce_available(&boot_cpu_data)) {
 		err = -EIO;
@@ -2580,22 +2584,13 @@ static __init int mcheck_init_device(void)
 				mcheck_cpu_starting, mce_cpu_down_dying);
 	if (err)
 		goto err_init_pool;
+	err = cpuhp_setup_state(CPUHP_AP_ONLINE_DYN, "x86/mce:online",
+				mce_cpu_online, NULL);
+	if (err < 0)
+		goto err_hp_online;
+	hp_online = err;
 
 	cpu_notifier_register_begin();
-	for_each_online_cpu(i) {
-		err = mce_device_create(i);
-		if (err) {
-			/*
-			 * Register notifier anyway (and do not unreg it) so
-			 * that we don't leave undeleted timers, see notifier
-			 * callback above.
-			 */
-			__register_hotcpu_notifier(&mce_cpu_notifier);
-			cpu_notifier_register_done();
-			goto err_device_create;
-		}
-	}
-
 	__register_hotcpu_notifier(&mce_cpu_notifier);
 	cpu_notifier_register_done();
 
@@ -2610,17 +2605,9 @@ static __init int mcheck_init_device(void)
 
 err_register:
 	unregister_syscore_ops(&mce_syscore_ops);
+	cpuhp_remove_state(hp_online);
 
-err_device_create:
-	/*
-	 * We didn't keep track of which devices were created above, but
-	 * even if we had, the set of online cpus might have changed.
-	 * Play safe and remove for every possible cpu, since
-	 * mce_device_remove() will do the right thing.
-	 */
-	for_each_possible_cpu(i)
-		mce_device_remove(i);
-
+err_hp_online:
 	cpuhp_remove_state(CPUHP_AP_X86_MCE_STARTING);
 
 err_init_pool:
-- 
2.10.2